Notes from the gap between AI demos and production.
I'm Josh Thomas. Chief AI Officer at Earthling Interactive, writing here about the integration, deployment, and developer-experience work that decides whether AI actually ships. No hype. Practitioner's notes.
AI ships when the unglamorous parts are solved.
Integration over models.
The hard part is never the model. It's the deployment pipeline, the error handling, the fallbacks, the eval loop. That's where shipping lives.
Developer experience first.
AI features fail when they're painful to build on. A good DX loop is the multiplier, for your team and for your users.
Demos lie, production tells the truth.
A demo runs the happy path on a good day. Production runs every path on every day. Design for the latter.
Compounding beats breakthrough.
The teams getting real value aren't chasing the frontier. They're compounding small, shipped improvements every week.
Get new posts in your inbox.
Occasional writing on AI integration, deployment, and the engineering reality behind the hype. Unsubscribe anytime.
Need help shipping AI at your organization? My team at Earthling Interactive does this work.